,,,.,
., ,
.-,
; ,
J. Am Ceram. Soc.. 75 131 507-13 (1992)
Volatility of Simulated High-Level Nuclear Waste Glass by
Thermogravimetric Analysis
John R. Harbour
Westinghouse Savannah River Company, Savannah River Site, Aiken, South Carolina 29808
River Company has detailed the strategy for demonstrating
compliance with the Waste Acceptance Preliminary Specifications in its Waste Form Compliance Plan2 for the Defense
Waste Processing Facility.
The Waste Acceptance Preliminary Specifications prohibit
foreign materials within the canistered waste forms. These
foreign materials include free gases, free liquids, explosives,
pyrophorics, combustibles, and organics. In addition, it is
specified that these foreign materials must not be released or
generated as a result of exposure of the canistered waste form
to the glass transition temperature (T,) of the waste glass. For
waste glass produced at the Defense Waste Processing Facility
at Savannah River Site, T, will be between 450" and 500°C.
These foreign materials can potentially exist either in the
waste glass itself or within the free volume of the canister.
(The canister free volume is that volume of the canister not
occupied by waste glass. Characterization of the gas within
the free volume of prototy ical canistered waste forms is the
subject of a separate study. ) This report focuses on the search
for any foreign materials which are released as the waste glass
is heated to T,. The technique of thermogravimetric analysis
(TGA) was used to detect volatility of the waste glass as it is
heated up to Tg. Experiments at temperatures several hundred degrees above T, were also carried out to determine the
nature of volatility as a function of temperature.
Reports on the volatility of waste glasses have previously
appeared in the literature. Gray4 has shown that volatility
from simulated high-level radioactive waste glasses is not significant below 1000°K. An extrapolation of his data to 500°C
(--Tg) gives a weight loss of 0.01 mg/m2 of glass surface.
Walmsley et al.' have also measured weight loss as a function
of temperature for a borosilicate glass. Extrapolation of their
data to 500°C gave a glass volatility of 0.24 mg/m2. Wilds6
measured weight loss from a sodium borosilicate glass containing simulated sludge. Wilds had to heat the glass to 900°C
for 3 h before a weight loss (0.02 mg/m2) could be detected.
Using a gamma radiation scanning method, Kamizono et aL7-'
detected 137Csand ln6Ru volatiles. This extremely sensitive
technique revealed that volatilities of these two radionuclides
were extremely low at 500°C. Actual detected values were
0.7 x lo-' mg/cm3 for 137Csand -0.1 x lo-' mg/cm3 for
lffiRu.Since sodium and lithium metaborates (see below) are
an order of magnitude more volatile than cesium, the volatilities of these species will be at least one order of magnitude
greater than the cesium volatility.
Hastie, Bonnell, and Plante1"311
used Knudsen effusion and
transpiration mass spectrometry to monitor glass volatility as
a function of temperature. They found that the volatile species generated were mainly sodium, lithium, and cesium
metaborates. This work has been confirmed by the mass spectrometric work of Asano and Yas~e,'~.''Asano and Kou,I4
Shultz et al., l5 and Wilds.16 Wenzel and SandersI7 used atomic
absorption spectroscopy to measure volatilities and found no
dependence of the volatility on relative humidity, a finding
which is consistent with the research of Hastie et a1.103"Terai
and KosakaI8 also observed metaborate volatiles between 900"
and 1300°C. Extrapolation of their volatility data to 500°C
gave
g/cm2/d.
The volatilities of simulated, high-level nuclear waste glasses
have been measured using thermogravimetric analysis
(TGA). These volatilities were measured in the region of the
glass transition temperature (T,) of the waste glasses, which
is between 450" and 500°C. These data were obtained because the Waste Acceptance Preliminary Specifications require that no foreign materials be released into the
canistered waste form upon heating of the canister to this
glass transition temperature. In fact, all of the waste glass
samples studied actually exhibited a net weight gain upon
heating. This weight gain was shown to be due to oxygen
uptake through oxidation of FeO. Powdered glass samples
did show a small weight loss which was smaller in magnitude than the weight gain and was associated with water
desorption. No true volatility was detected to the level of sensitivity (0.01 wt%) of the TGA instrument. This converts to a
sensitivity of 330 pg/mz of glass surface and a corresponding minimum value of 41 mg of volatiles for each Defense
Waste Processing Facility canistered waste form. TGA experiments carried out at higher temperatures (800°C) revealed
that organic concentrations in the waste glasses are less
than 0.01 wt%. Thus, these results demonstrate that the Defense Waste Processing Facility will be able to comply with
the Waste Acceptance Preliminary Specifications on the exclusion of foreign materials from the canistered waste forms,
after exposure to T,. [Key words: glass, nuclear waste,
volatility, thermogravimetry, models.]
P
I. Introduction
T
Department of Energy has 130 million liters of highlevel nuclear waste stored in carbon steel tanks at its
Savannah River Site. The Defense Waste Processing Facility
has been constructed in order to convert this alkaline slurry
into a durable borosilicate glass which is poured into a stainless steel canister and sealed. The sealed canister is then transported to the Glass Waste Storage Building at the Savannah
River Site for temporary storage prior to transfer to the Department of Energy's Office of Civilian Radioactive Waste
Management for shipment to the federal repository.
The Office of Civilian Radioactive Waste Management established a Waste Acceptance Committee to develop a process
for acceptance of high-level nuclear waste from Department
of Energy defense facilities. As part of this process, this committee produced the Waste Acceptance Preliminary Specifications,' which detail the requirements that must be met
before Office of Civilian Radioactive Waste Management will
accept the canistered waste forms. Westinghouse Savannah
HE
C. M. Jantzen-contributing
editor
Manuscript No. 196637. Received June 17, 1991; approved November 12,
1991.
Prepared in connection with work done under Contract No. DE-AC0989SR18035 with the U.S. Department of Energy. By acceptance of this
paper, the publisher and/or recipient acknowledges the U.S. Government's
right to retain a nonexclusive, royalty-free license in and to any copyright
covering this paper, along with the right to re roduce and to authorize
others to reproduce all or part of the copyrighte8paper.
507
508
Journal of the American Ceramic Society -Harbour
It is clear from the above studies that the volatilities of simulated high-level nuclear waste borosilicate glasses are very low
and that temperatures of -800°C are required to achieve detectable levels of volatiles (except for the work of Kamizono
et ~ l . ' - ~ Extrapolated
).
volatilities at Tg,(-5OO"C), estimated
from the data presented above, are very small but vary over
several orders of magnitude. However, the Waste Acceptance
Preliminary Specifications require volatility data at Tg,from
Defense Waste Processing Facility waste glass. TGA experiments were therefore carried out on Defense Waste Processing Facility simulated high-level nuclear waste glasses
produced at the Savannah River Site in a temperature range
which includes Tg.This report details the results of these experiments and provides evidence for a lack of glass volatility
(to the sensitivity of the TGA instrument) at Tg.
11.
Experimental Procedure
TGA experiments were carried out using a Du Pont 951
thermogravimetric analyzer and 2100 thermal analyzer.
Samples (-100 mg) were placed in platinum boats. The purge
gas was either helium, oxygen, or air. The flow rate was normally 0.4 standard cubic feet per hour (1 ft3 -2.83 x lo-' m3.)
The samples of simulated waste glass were subjected to two
different temperature cycles, profile 1 (to 500°C) and profile 2
(to 800°C).
In the temperature profile referred to as profile 1, the
sample is heated to 450°C at a rate of 10"C/min. It is held at
450°C for 10 min and then slowly ramped at a rate of
O.S"C/min to 500°C. After a 10-min holding period at 500"C,
the sample is cooled to room temperature using the reverse of
the profile used for sample heating. The objective here was to
ensure that the glass sample was heated within the region of
Tgfor at least 4 h.
In temperature profile 2, the sample is ramped from room
temperature to 800°C at a rate of 10"C/min (most of these
sample runs included a 3-h hold at 60°C) and then held at
800°C for 10 min before being cooled at a rate of lO"C/min.
The profile is then repeated (without the 3-h hold period).
Glass samples were obtained from canisters which had been
filled during large-scale pilot plant campaigns, referred to as
scale glass melter (SGM) campaigns, at the Savannah River
Site with one exception. The glass sample from SGM 8 was
intercepted between the melter pour spout and the canister
nozzle. The SGM campaigns used a variety of simulated
borosilicate waste g l a s ~ e s . 'The
~ eighth SGM campaign
(SGM 8) used black frit 165, which contains simulated sludge
but no precipitate hydrolysis aqueous product simulant. All of
the other campaigns used a simulated glass frit/sludge/precipitate hydrolysis aqueous feed spiked with various levels of organics to cover the expected range of organics which will be
present during the running of the Defense Waste Processing
Facility.
Black frit 165 contains the following component^,'^ in wt%:
Si02 (55), B z 0 3(7.2), LizO (5.0), NazO (11), MgO (0.7), CaO
(1.4), Zr02 (0.7), Fez03 (11.3), A1203 (4.3), MnOz (2.5), and
NiO (0.9). A typical coupled-feed glass" containing 65.4%
frit, 26.6% sludge, and 8.0% precipitate hydrolysis aqueous
feed contains the following components, again in wt%: Si02
(49.1), B2O3 (lO.l), Liz0 (3.2), N a20 (lO.l), MgO (1.3), CaO
(0.9), Fez03 (11.7), A1203(4.7), Mn02 (3.2), NiO (l.O), Ti02
(1.3), K 2 0 (3.2), C s 2 0 (0.06), and SrO (0.05).
The glass samples were analyzed with the thermogravimetric analyzer as both glass pieces and glass powders. The glass
powders were created by milling the larger glass pieces, using
a mechanical grinder (Tekmar grinding mill with a tungsten
carbide blade), and then collecting only those particles which
passed through a 200-mesh screen. Glass pieces were fragments less than 100 mg. Typically, from one to five glass
pieces were used to give a loading of -100 mg.
Calcium oxalate monohydrate from Du Pont was used as
the "standard for TGA experiments. Calcium oxalate mono-
Vol. 75, No. 3
hydrate provides a reference to show that the instrument is
operating effectively and checks the stability of the instrument over time. This material was run at the beginning, end,
and once during the middle of data acquisition. The TGA
curve showed the expected weight losses within experimental
error for this material. The values determined were 12%,
18%, and 31% for the three different weight loss regions. The
corresponding weight losses reported by Du Pont2' are 12.5%,
18.5%, and 30.3%.
The temperature ranges of the three weight losses using
calcium oxalate monohydrate also provide a general indication of proper temperature calibration. The most important
temperature range for the purpose of this study includes the
glass transition temperature, Tg, of the glass, which is between 450" and 500°C. This corresponds to the second temperature range for calcium oxalate monohydrate, which has
been reported in the literature as 405" to 523"C,20400" to
500°C,21and 400" to 500"C.22 This can be compared to the
range observed in these experiments of 390" to 490°C. This
indicates that the temperature reported on the TGA curves
for this work is -10" to 20°C lower than the actual temperature. The third temperature range for these experiments was
significantly less than reported by others. The range for this
study was 565" to 700°C versus reported values of 626" to
793"C,2' 660" to 830"C,*' and 620" to 850"C.22This indicates
that for TGA curves using profile 2, the temperature indicated in the plot underestimates the actual temperature by
-100°C. The data are presented in this paper without correcting for this divergence, since it does not alter the conclusions
of this work regarding volatility.
The TGA curves obtained from calcium oxalate monohydrate at the beginning and near the middle of the experimental period were essentially identical. The calcium oxalate
monohydrate run at the end, however, showed a second temperature range of 400" to 500°C and a third range of 610" to
725°C. These changes reduced the difference between actual
and measured temperatures discussed above.
In terms of absolute values of weight, the TGA underestimates the true value by 4.1%. This was determined using a
calibrated set of weights. Hence, an observed weight change in
these experiments of 400 pg translates to an actual change of
384 pg. Since this correction is close to the detection limit
of the apparatus, no corrections were made to the weight
changes recorded in this report.
The scanning electron microscopy data were obtained on an
IS1 Model DS-130 scanning electron microscope. Polystyrene
spheres of diameter equal to 29.64 ? .06 pm were added to
the glass samples as an internal standard. The spheres were
obtained from NIST, as SRM 1961.
A Leeds and Northrup Microtrac 11, model 158704 laser
light scattering instrument was used to obtain the particle
size distribution. The glass powder was dispersed (-0.3 g of
glass in 25 mL) in a 0.6 vol% NazHPO4aqueous solution (distilled water). This was subsequently diluted in distilled water
(-25 mL) and placed in the sample cell. The "standard used
was an alumina powder, whose particle size distribution had
not changed significantly for the past few months.
The BET data were acquired using a Digisorb 2600 BET
adsorption surface area analyzer by Micromeritics. A "standa r d sample of alumina was used and a surface area of
0.468 m2/g obtained. The expected value for the alumina is
0.47 2 0.02.
The Fe2+/Fe3+ratios were determined colorimetrically
using the procedure developed by B a ~ m a n n . 'The
~ experiments were performed using a Hewlett-Packard 8451 diode
array spectrophotometer.
111.
Results
(1) Powder Characterization
This TGA study of simulated waste glasses used two types
of samples. The first type, referred to as glass pieces, was
March 1992
Volatility of Simulated High-Level Nuclear Waste Glass
small fragments of glass which had been removed from the
canisters. The other type of glass sample was powder. This
milling of the larger glass pieces into a powder was done in
order to increase the surface area of the glass sample. This is
important because volatilization from the surface is the expected mechanism for weight loss at temperatures near Tg.
The milling produced a polydisperse powder, and only that
portion of the powder which passed through a 200-mesh filter
was collected. Figure 1 shows a scanning electron micrograph
of the milled glass containing some added -30-pm latex
spheres for reference. A determination of the particle size distribution by laser light scattering revealed a mean diameter of
the volume distribution that ranged from 43 to 60 pm. The
BET surface areas ranged from 0.23 to 0.39 m'/g.
(2) TGA Results
A typical TGA curve for a simulated waste glass (produced
during the seventh SGM campaign) subjected to profile 1 (see
Experimental Procedure section) is shown in Fig. 2. The
curve shows the change in weight of the sample, in weight
percent, as a function of temperature. This particular powdered sample of -100 mg was purged with oxygen during the
heat treatment. From this curve, it is evident that the glass
sample has actually gained weight as a result of this heat treatment. The apparent weight gain (measured after the sample
has cooled back down to 25°C) is -430 pg or 0.43 wt%.
Figure 3 gives the same data except that the change in weight
is plotted against time.
The slope of the cooling portion of this curve (3 x
wt%/OC) is the baseline change with temperature and is
reproducible (i.e., if the sample is subjected to the same temperature profile a second time, the curve will follow this
slope during both heating and cooling). This implies that
(1) no further changes are occurring upon additional heat
treatment and (2) the baseline is not flat.
Since the initial slope for the heating cycle is less than this
baseline slope, the data also reveal that this particular sample
must be losing weight during the initial heating of this powdered glass sample.
Figure 4 shows the type of curve obtained using profile 2
for a powdered sample of SGM 7 with oxygen purging. The
weight gain (-570 pg) has plateaued by -650°C, with the
repeat cycle demonstrating that no additional weight change
has occurred. This plot reveals that the scatter in weight after
reaching saturation is -10 pg (0.01%) as measured after the
sample has returned to room temperature. An initial weight
loss is also evident with profile 2, since the slope of the initial
heating curve is less than the slope of the cooling curve.
I
100.6
I
c
100.5
100.4
Z
509
I
I
/
-
I-I
I
I
t
100.2
0
l
300
200
100
400
I
i
500
Temperature, "C
Fig. 2. Profile 1, TGA curve of SGM 7 powdered glass, using oxygen as a purge gas.
Although the data shown in Fig. 4 were from a powdered
sample, the powder softens at the higher temperatures of
profile 2, causing the particles to coalesce into a thin film of
glass in the platinum boat. This is not the case for profile 1,
where the maximum temperature reached is only slightly
above the glass transition temperature of the glass. Hence,
the sample is still a powder after being subjected to profile 1,
although some clumping of the particles was observed. This is
probably due to the fact that the glass had reached the softening temperature. It was easily redispersed.
Three sample loadings of powdered SGM 7, purged with
oxygen, were run using profile 2 which revealed that the TGA
results, in terms of percent weight loss, are not significantly
dependent on the amount of material in the platinum boat.
(3) Effect of Purge Gas
The effect of purge gas on the overall weight gain is significant. For example, using profile 2, the powdered glass sample
from SGM 7 showed a weight gain of -300 pg with helium, a
gain of 440 pg with air, and a gain of 540 pg with oxygen.
This same general trend is also evident using profile 1. The
flow rate of purge gas also has an effect on the apparent
weight gain. At higher helium flow rates of 3 standard cubic
feet p e r hour (compared t o t h e normal f l o w rate of
0.4 standard cubic feet per hour) the weight gain was further
reduced. These results indicate that it is the oxygen in the
purge gas that leads to the observed weight gain. It is clear
from this study that even with helium purging, significant
oxygen uptake can occur. This is due, at least in part, to insufficient removal of oxygen within the TGA apparatus as
I
100.6
i
100.4
100.0
0
50
100
150
200
250
300
350
Time, min
Fig. 1. Scanning electron micrograph of SGM 7 glass powder. The
spheres are an internal standard of polystyrene latex of -30-pm
diameter.
Fig. 3. Profile 1, TGA curve of SGM 7 powdered glass, using oxygen as a purge gas and plotted as percent weight change versus
time. Same data as in Fig. 2.
510
Journal of the American Ceramic Society -Harbour
100.8
Vol. 75, No. 3
Table I. Weight Gain as a Function of the SGM Campaign
for Powders Using Profile 1
SGM
100.6
campaign
Weight gain (wt%)
5
6
7
8
9
10
0.22
-0.1
0.33
-0.09
0.25
0.30
profile 1, large pieces of glass from SGM 7, for example, re-
99.8
vealed an overall weight loss of -15 pg, using helium as the
8
purge gas, compared to the control experiment in which the
0
200
400
600
800
Temperature, "C
Fig. 4. Profile 2, TGA curve of SGM 7 powdered glass, using oxygen as a purge gas.
reflected by the fact that higher flow rates of helium reduced
the weight gains.
There is an additional change which occurs with the
amount of oxygen in the purge gas. A small peak near 400°C
(see Fig. 5) in the heating portion of the cycle is clearly evident in air but much less pronounced in either helium or oxygen. The origin of this peak, which is reproducible and
present in all glasses studied, is unknown.
(4) Effect of Waste Glass Composition
The TGA results on waste glasses from six different SGM
runs are presented in Table I. Results were obtained with helium purging using profile 1 at a flow rate of 0.4 standard
cubic feet per hour. The reported numbers are the differences
between the final and initial weights. For profile 1, samples
from campaigns 6 and 8 showed overall weight losses of 100
and 90 pg, respectively, with heat treatment, while the other
four samples all showed gains from 220 to 330 pg.
The results from profile 2 (Table 11) demonstrate that all
6 of the waste glass samples gained weight. As in profile 1,
however, glass samples from campaigns 6 and 8 formed a
unique subset, which showed the least amount of weight gain.
The facts that SGM 6 and SGM 8 glasses gained weight using
profile 2, but lost weight with profile 1, resulted from exposure to a higher temperature and the use of air for purging
with profile 2.
(5) Effect of Surface Area
The surface area of the glass sample used has a significant
effect on the observed weight changes for both profiles. For
100 8
8W
100.6
64
E
P 100.4
g
6ooc
1
empty boat showed a weight loss of -20 pg. Within experimental error, then, no weight change was observed for the
glass pieces. This is contrasted to the powdered form of the
same glass, which showed an overall weight gain of -400 pg
under similar conditions. In addition, the weight loss observed in the initial heating cycle of profile 1with the powder
is not observed with the glass pieces.
Table I1 gives the results of the powdered samples versus
glass pieces using profile 2. In all cases, the weight gain by
the pieces was less than that for the powders. To ensure that
saturation of the weight gain had occurred with the glass
pieces, SGM 7 glass pieces were subjected to a 30-h heat
treatment. No further weight gain was observed during the
repeat cycle. It was noted that weight gain with the pieces did
not begin until 500"C, with most of the gain occurring after
the temperature reached 800°C.
With the glass pieces, the initial slope during heating was
the same as the final slope during cooling for profile 2. This
is in contrast to the powders, where the initial slope is always
less than the final slope (an indication of a weight loss).
Hence, the pieces of glass, to the sensitivity of this instrument, do not show any weight loss with either profile.
(6) Sensitivity of TGA Measurement
The sensitivity of the TGA measurement is defined here as
the minimum change in weight which can be determined for
these 100-mg samples and is reported in terms of weight percent. For the powders, the sensitivity is also expressed in
terms of minimum change in weight per unit surface area.
Sensitivity was determined for both powders and pieces using
both profiles. In certain cases, sudden changes in the weight
greater than the reported sensitivity occurred during the
many hours that are required to obtain a TGA curve. However, these changes are random and associated with electrical
or mechanical noise and are identifiable by instantaneous
change. This noise was reduced by acquiring the data in
overnight runs.
For glass powders, the time dependence of the weight
change for a profile 1 heat treatment revealed that the weight
gain reached a limiting value by the 200-min mark in the 450"
to 500°C temperature range. After this limiting value and during the repeat cycle of this profile, the TGA curve simply
followed the baseline within a scatter range of -10 pg for the
100-mg samples. Therefore, after the initial weight loss and
f
4ooF
Table 11. Weight Gain as a Function of
Surface Area Using Profile 2
Weight gain (wt%)
SGM
carnDaien
5
6
7
8
9
10
Powder
Pieces
0.46
0.13
0.45
0.16
0.45
0.39
0.33
No change
0.30
0.12
0.26
0.22
Volatility of Simulated High-Level Nuclear Waste Glass
March 1992
the subsequent weight gain, no additional change of weight
for the glass was observed in excess of 0.01 wt%.
For the glass pieces or powder subjected to the heat treatment of profile 2, a scatter of -10 pg was observed after the
weight gain reached saturation. Hence, in all cases, the minimum change detected was 10 pg, corresponding to a sensitivity of 0.01 wt%.
Near the glass transition temperature, volatility is expected
to be a surface-controlled property, and therefore the glass
volatility in terms of surface area becomes important. Since
the BET surface area of the powders is -0.3 m2/g, the sensitivity with the powders is 330 pg/m2.
(7) TGA of Glass Frits
Black frit 165, which was the simulated waste glass feed
used in SGM-8, showed a TGA curve of the type shown in
Fig. 2 with a weight gain of -150 pg. The corresponding 165
white frit, which does not contain iron, showed no weight
gain from the TGA curve. An Apec frit, a Cataphote frit, and
a Ferro 202, -200 mesh glass frit (these frits contained
-77 wt% S O 2 , -8 wt% BzO3, -7 wt% LizO, and -6 wt%
Na20), all essentially void of iron, showed no weight gain in
the TGA experiments but did reveal weight losses of
-0.10 wt% (-100 pg), -0.04 wt% (-40 pg), and -0.2 wt%
(-200 pg), respectively. These results reveal that iron must
be present within the glass or glass frit in order to observe a
weight gain.
IV. Discussion
(I) Weight Gain
Weight gain, the dominant feature in the TGA curves of
waste glass samples, correlates with the amount of oxygen in
the purge stream and the presence of iron in the glass. It
therefore appears that the weight gain is due to oxygen uptake. Iron is present in these glass samples in both a reduced
form (FeO) and in an oxidized form (Fe2O3). The presence of
FeO provides a means by which uptake of oxygen can occur:
4FeO
+ O2+.
2FeZO3
If this mechanism accounts for the observed weight gain,
then the ratio of Fe2+/Fe3+in the waste glasses should reflect
this. Table 111 summarizes the Fe2+/Fe3+ratios for the waste
glasses of this study both before and after heat treatment during the TGA runs. This table reveals that the waste glasses
from SGM 6 and SGM 8 have the lowest ratios of Fe2+/Fe3+,
i.e., are the most oxidized. This implies that there is far less
FeO present in the SGM 6 and 8 samples available to react
with oxygen. The measured oxygen uptake values (weight
gain) for these samples (Table 111) are consistent with this
ratio. The other samples, with higher ratios, all exhibit significant oxygen uptake.
The values presented in Table 111 were obtained using
profile 1 under different conditions. Thus, SGM 6 and 8 glass
powders were run up to 500°C in a helium atmosphere, and
no weight gain was observed. SGM 9 and 7 glass powders
were run in oxygen to maximize the weight gain. The SGM 5
powder was run in helium.
From the weight gain by TGA (measured by adding 100 pg
to the difference between final and initial weights (see Weight
Loss section) and the initial and final ratios of Fe2+/Fe3+,one
can calculate the amount of elemental Fe in the glass. These
Table 111. FeZ+/Fe3+Ratios in SGM Waste Glass Powders
SGM
camoaien
5
6
7
8
9
Fe2+/Fe3+ratio
Initial
Final
0.72
0.24
1.70
0.22
0.74
0.08
0.21
0.25
0.19
0.00
Weight gain
(UP)
wt% Fe
350
None
550
None
440
7.3
9.0
7.3
511
calculated values are shown in the last column of Table 111.
These values are close to the amounts of Fe initially fed to
the glass melter. For example, the Fe introduced into the
melter for SGM 9 was -7.7 wt%. Hence, the weight gain observed in these experiments can be accounted for, to first
order, b the amount of Fez+ in these glass samples. (The
Fe2+/Feyi ratios obtained from the samples heated in the
TGA instrument are only approximate. This is due to the limited amount (-100 mg) of sample available for each analysis.)
The waste glass powders from the six SGM runs were
either greenish gray or yellowish brown. The yellowish brown
samples from SGM 6 and 8 showed no color change after
being heated through profile 1. The other SGM powders were
greenish gray in appearance and underwent a color change to
yellow to dark brown after being heated through profile 1.
Hence, the increase in weight is correlated with a color
change within the glass. Since Fe2+ is green whereas Fe3+ is
rust colored, these color changes are also consistent with
reaction (1).
The difference in weight gain as a function of surface area
can be rationalized in terms of oxygen diffusion in the particles. For profile 1, very little oxygen uptake occurs with the
glass pieces, whereas close to the maximum value of oxygen
uptake can be observed with the higher surface area powders.
This demonstrates that, even at temperatures near Tg,oxygen
can diffuse within the bulk of the glass, provided the particle
size is small enough. The light scattering and scanning
electron micrograph data reveal that the powder consists of
polydisperse, irregularly shaped particles with platelike morphology. This implies that the average diffusion length for
oxygen under these conditions of time and temperature is
much less that the average volumetric particle size of 43 pm.
Half the minimum dimension of each particle is the maximum oxygen diffusion length necessary for oxygen to reach
FeO. This dimension, although difficult to determine, appears to be less than 10 pm from the scanning electron micrograph data for the larger particles of the powder.
For profile 2, the glass pieces eventually showed a weight
gain which was approximately half that of the powders. During this heat treatment, the glass flowed to form a film on the
platinum boat. In spite of the fact that the temperature profile
was higher than profile 1, the diffusion of additional oxygen
within this film was limited under these conditions of time
and temperature, as evidenced by a smaller weight gain than
the powders. This was true even though the glass film was
held at 800°C for 20 h.
(2) Weight Loss
The TGA curves for the powdered glass samples showed
slopes in the initial heating cycle which were less than the
slope of the baseline. This implies that weight loss is occurring during the initial heating of these powdered samples. On
the other hand no weight loss, within experimental error, was
observed with the larger pieces of glass.
The magnitude of the weight loss with the glass powders
can be estimated from the results of several experiments.
Figures 2 and 4 reveal the initial weight loss for SGM 7 glass
powder using both profiles 1 and 2. The weight loss begins
below 100°C and ends between 200" and 300°C. Extrapolating
back to the ordinate using the normal baseline (cooling slope
of these curves) gives an intersection near the 0.1 wt% loss
region (a loss of 100 pg for the 100-mg sample) for both of
these curves. Experiments with empty platinum boats revealed that the weight loss was approximately 20 p g
(0.02 wt%) under these conditions.
A second way to estimate the weight loss is to consider the
results of SGM 6 and 8 glass powders heated using profile 1 in
helium. Under these conditions, weight gain is essentially precluded and only weight loss is observed. These oxidized powders showed an overall weight loss of 0.1 wt% for SGM 6 and
0.09 wt% for SGM 8. This corresponds to losses of 100 and
90 pg, respectively.
Journal of the American Ceramic Society -Harbour
512
Finally, TGA experiments using glass frit samples revealed
a weight loss in the range from 0 to 0.2 wt%. These glass frit
samples contained no iron and consequently exhibited no
weight gain. Therefore, the only change that occurred was
due to weight loss.
The conclusions from these three experiments are that the
weight loss with the glass powders is -0.1 wt% and that it
occurs at relatively low temperatures (less than 300°C).
This weight loss with the powders can be explained as follows. The glass powder has a surface area 600 times greater
m2/g). Durthan the surface area of the glass pieces (5 x
ing the milling process for producing the powder, water from
the atmosphere can adsorb onto the newly formed surfaces of
the particles (reaction (2)). The amount of water adsorbed is
directly proportional to the specific surface area of the powder. A weight loss is then observed in the TGA experiments
when the powder is heated (reaction (3)). This weight loss is
+
H20
powder H 2 0 + (powder)adsorbed
(pOWder)adsorbedH 2 0 + heat + powder
(2)
+ H20
(3)
due to the desorption of the adsorbed water. The fact that the
weight loss occurs at relatively low temperatures (see Figs. 2
and 4) is also consistent with desorption of weakly adsorbed
water from the glass surface. In terms of surface area, the
desorption of adsorbed water from the surface of the glass
pieces should be 600 times less than this, or 0.17 pg for a
100-mg sample. This is well below the sensitivity of the TGA
measurement.
The weight loss due to water desorption and the weight
gain due to oxygen uptake have the potential to mask the
true volatility of the waste glass. However, water desorption
occurred at temperatures between 50” and 300°C and consequently did not interfere with measurements near Tg. Also,
certain powdered samples could be run under conditions
where no weight gain was observed. In these cases no volatility was observed at T g . For those samples which gained
weight, no further changes in weight due to volatility were
observed after weight gain due to oxygen uptake had stopped.
These results suggest that masking does not alter the conclusion that glass volatility is less than 0.01 wt% at 7’’.
It was demonstrated that the presence of oxygen does not
effect the volatility of the glass samples. That is, the volatility,
whatever its value, is still less than the detection level, independent of the oxygen concentration in the purge gas. Since
the compressed air used in these experiments has a relative
humidity of -30%, the same arguments can be made for the
lack of an effect of water on the glass volatility for a relative
humidity value of 30% (both the helium and oxygen are dry).
The relative humidity values in the canisters are expected to
be less than this value: while the oxygen level will most likely
be close to that of air (-20%).
(3) Defense Waste Processing Facility Canister Volatility
The fact that glass volatility is less than 0.01 wt% or
330 pg/m2 can be used to estimate an upper limit of the
volatility within a Defense Waste Processing Facility canister
heated to Tg.A monolith of glass (a cylinder of 2.4-m height
with a radius of 0.3 m) weighing 1700 kg inside a Defense
Waste Processing Facility canister has a geometric surface
area of -5 m2. Due to cracking of the glass,24 this surface
area can increase by a factor of 25, giving a total surface area
of glass within the canister of 125 m2. Therefore these TGA
experiments give an upper limit for volatility of 41.3 mg for a
single canister, at Tg.
V. Conclusions
This work has demonstrated that simulated high-level waste
glass uptakes oxygen during heat treatment. This oxygen uptake is explained completely by the oxidation of FeO to
Fe203.Hence the weight gain is controlled by the amount of
Vol. 75, No. 3
FeO in the glass and the availability of oxygen. For these
SGM waste glass samples, the maximum weight gain was
-0.55 wt%. With glass powders, there is significant uptake in
the temperature range from 200” to 500°C. With glass pieces,
significant oxygen uptake begins near 500”C, with most of the
weight gain occurring at temperatures near 800°C. These differences in oxygen uptake between the powders and the
pieces can be explained by the oxygen diffusion length at
these temperatures.
These same waste glass samples also exhibit a weight loss if
the sample is in powdered form. This loss is most likely due
to the desorption of H 2 0 which had adsorbed onto the newly
created surface during the milling of the glass into the powder. The loss is estimated to be between 0.05 and 0.1 wt%,
which corresponds roughly to 0.5 to 1 water molecule/A*.
~
true
All of the scale glass melter glass ~ a m p l e s ’showed
volatilities less than 0.01 wt%. This is equivalent to a maximum weight loss through volatilization of 330 pg/m2 of glass
surface. For a canister, this translates into an upper limit for
the volatility of 41.3 mg released within a single canister as a
result of heating the canister to Tg.Hence, it may be concluded that any free liquids, free gases, combustibles, pyrophorics, and explosives released upon heating the waste glass
to Tgwere also less than the detection limit of 330 pg/m2 of
glass surface, or 41 mg/canister.
The lack of organics within the waste glass was demonstrated by the fact that heating the SGM waste glass powders
and pieces (containing different amounts of organics in the
feed) in oxygen up to 800°C did not give rise to detectable
volatility. Under these conditions, organic compounds would
burn to carbon dioxide and other volatile compounds, which
would lead to an observable weight loss. This limits the
amount of organics to be less than or equal to the detection
limit of 0.01 wt%.
These results will be used to demonstrate that the Defense
Waste Processing Facility is in compliance with the Waste Acceptance Preliminary Specifications.
Acknowled ments:
All the data presented here on TGA were acquired
J. Kerrigan provided guidance in the acquisition and interby A. B. Ra
pretation ofY;he TGA data. D. F. Steedly carried out the SEM work, while
J.T. Hunter determined the particle size distribution. R. A. Malstrom measured the BET surface areas. Glass sam les were obtained from C.M.
Jantzen, N. E. Bibler, W.G. Ramsey, a n 2 R. F. Schumacher. D. C. Beam
generated the powdered lass samples. C. M. Jantzen pointed out the mechanism given in Eq. (1) ancf W. G. Ramsey su gested measuring the Fe2+/Fe3+
ratio, which was subsequently performed %y N. S. Wallace. M. J. Plodinec
provided guidance throughout this work.
6’.
References
!Office of Civilian Radioactive Waste Management, “Waste Acceptance
Preliminary Specifications of the Defense Waste Processing Facility HighLevel Waste Form,” U.S. DOE Report No. OGR/B-8, Rev. 1, February 1988.
2U.S. Department of Energy, Savannah River Site, “Defense Waste Processing Facility Waste Form Compliance Plan,” U.S. D O E Report
No. WSRC-SW4-6, March 1990.
’J. R. Harbour, T. J. Miller, and M. J. Whitaker, “Exclusion of Foreign
Materials from SRS Canistered Waste Forms: Characterization of the Gas
within the Free Volume”; in Proceedings of the 1991 Infernational High-Level
Radioactive Waste Management Conference, Las Vegas, NV, 1991. American
Nuclear Society, LaGrange Park, IL and American Society of Civil Engineers, New York.
‘W. J. Gray, “Volatility of a Zinc Borosilicate Glass Containing Simulated
High-Level Radioactive Waste,” U.S. DOE Report No. BNWL-2111, 1976;
W. J. Gray, “Volatility of Some Potential High-Level Radioactive Waste
Forms,” Radioact. Waste Manage., 1, 147 (1989).
5D. Walmsley, B. A. Sammons, and J. R. Grover, “Volatility Studies of
Glasses for the FINGAL Process,” Report No. AERE-R 5777. Atomic Energy Research Establishment, Harwell, U.K., July 1969.
%.W. Wilds, “Vaporization of Semi-Volatile Components from Savannah
River Plant Waste Glass,’’ U.S. DOE Report No. DP-1504, 1978.
’H. Kamizono, S. Kikkawa, Y. Togashi, and S. Tashiro, “Volatilization of
‘”Cs and Io6Ru from Borosilicate Glass Containing Actual High-Level
Waste,”J. Am. Ceram. Soc., 72 [8] 1438-41 (1989).
*H. Kamizono, S. Kikkawa, S. Tashiro, and H. Nakamura, “Volatilization
of Cesium from Nuclear Waste in a Canister,” Nucl. Technol., 72, 84-88
(1986).
9H. Kamizono, S. Kikkawa, S. Tashiro, N. Nakamura, and H. Kanazawa,
“Air Contamination by Cesium in a Canister Containing Nuclear Waste
Glass,”J. Nucl. Muter., 149, 113-16 (1987).
March 1992
Volatility of Simulated High-Level Nuclear Waste Glass
IoD.W.Bonnell, E. R. Plante, and J.W. Hastie, “Vaporization of Simulated
Nuclear Waste Glass,” J. Non-Cryst. Solids, 84, 268-75 (1986).
“J.W. Hastie, E. R. Plante, and D.W. Bonnell, “Vaporization of Simulated
Nuclear Waste Glass,” Report No. NBSIR 83-2731. National Bureau of Standards, Washington, DC.
”M. Asano and Y. Yasue, “Mass Spectrometric Study of the Vaporization
of Sodium Borosilicate Glasses,” J. Nucl. Muter., 138, 65-72 (1986).
I’MM.Asaco and Y. Yasue, “Vapor Species over Na20-B2O3-SiO2-Cs2O
Glass,”J. Nucl. Sci. Technol., 22, 1029-32 (1985).
I4M. Asano and T. Kou, “Vaporization of Sodium-Rubidium Borosilicate
Glasses,”Pkys. Chem. Glasses, 30, 39-45 (1989).
ISM. M. Shultz, V. L. Stolyarova, and G . A. Semenov, “Mass Spectrometric Study of Thermodynamic Properties and Vaporization Processes in the
Na~O-B20~.-GeO2Glass-Forming Melts,” J. Non-Crysl. Solids, 38-39, 58186 (1980).
I6G.W. Wilds, “Vaporization of Semi-Volatile Components from Savannah
River Plant Waste Glass,” U.S. DOE Report No. DP-1504, 1978.
”3.T. Wenzel and D. M. Sanders, “Sodium and Boron Vaporisation from a
513
Boric Oxide and a Borosilicate Glass Melt,” Phys. Chem. Glasses, 23, 47-52
(1982).
Terai and E. Kosaka, “Volatilization of Low Temperature Borosilicate Glasses for High-Level Radioactive Wastes at Elevated Temperatures,” Osaka Kogyo Gijutsu Shikensho Kiho, 21, 150-56 (1976).
’”C. M. Jantzen, “Glass Compositions and Frit Formulations Developed
for Defense Waste Processing Facility,” U.S. DOE Report No. DPST-88-952,
1988.
’““Instruction Manual, Du Pont 950 Thermogravimetric Analyzer,”
Manual No. G-125. E. 1. du Pont de Ncmours & Co., Wilmington, DE.
z‘C. Duval, Inorganic Thermogravimetric Anolysis, 2nd and revised ed.;
p. 281. Elsevier Publishing Co., Barking, Essex, U . K . , 1963.
’*A. E. Newkirk, “Thermogravimetric Measurements,” Anal. Chem., 32,
1558 (1960).
23E.W.Baumann, “Colorimetric Determination of Ferrous/Ferric Ratio in
Glass,” U.S. DOE Report No. DPST-87-304, 1987.
24M.J. Plodinec, “Glass Fracture and Defense Waste Processing Facility
Product Performance,” US. DOE Report No. DPST-89.220, 1989.
0